1. Field of the Invention
The present invention relates to a metal(M)-insulator(I)-semiconductor(S) device generally known as a MIS semiconductor device (also known as an insulated-gate semiconductor device). Such MIS semiconductor devices include, for example, MOS transistors, thin-film transistors, and the like.
2. Description of the Prior Art
In the prior art, MIS semiconductor devices have been fabricated using self-alignment techniques. According to such techniques, a gate electrode is formed on a semiconductor substrate or a semiconductor film with a gate insulating film interposed therebetween, and using the gate electrode as a mask, impurities are introduced into the semiconductor substrate or the semiconductor film. Thermal diffusion, ion implantation, plasma doping, and laser doping are typical methods of introducing impurities. With self-alignment techniques, the edges of the impurity doped regions (source and drain) can be substantially aligned with the edges of the gate electrode, eliminating the overlap between the gate electrode and the doped regions (that could give rise to the formation of parasitic capacitances) as well as the offset that causes separation between the gate electrode and the doped regions (that could reduce effective mobility).
The prior art process, however, has had the problem that the spatial carrier concentration gradient between the doped regions and their adjacent active region (channel forming region) formed below the gate electrode is too steep, thus causing an extremely high electric field and increasing, in particular, the leakage current (OFF current) when a reverse bias is applied to the gate electrode.
To address such a problem, the present inventor et al. have found that an improvement can be made by slightly offsetting the gate electrode with respect to the doped regions, and also that an offset of 300 nm or less can be obtained with good reproducibility by forming the gate electrode from an anodizable material and by introducing impurities by using the resulting anodic oxide film also as a mask.
Furthermore, in the case of ion implantation, plasma doping and other methods that involve driving high-velocity ions into a semiconductor substrate or a semiconductor film to introduce impurities, the crystallinity of the semiconductor substrate or film needs to be improved (activation) since the crystallinity of the structure where ions are driven is damaged by the penetrating ions. In the prior art, it has been practiced to improve the crystallinity by thermal means using temperatures of 600.degree. C. or higher, but according to the recent trend, lower process temperatures are demanded. In this view, the present inventor et al. have also shown that the activation can be accomplished by using laser or equivalent high-intensity light and that such activation has significant advantages for mass production.
FIG. 2 shows a fabrication process sequence for a thin-film transistor based on the above concept. First, a base insulating film 202 is deposited over a substrate 210, and a crystalline semiconductor region 203 is formed in the shape as an island, over which an insulating film 204 that acts as a gate insulating film is formed. Then, a gate connection 205 is formed using an anodizable material. (FIG. 2(A))
Next, the gate connection is anodized to form an anodic oxide film 206 to a thickness of 300 nm or less, preferably 250 nm or less, on the surface of the gate connection. Using this anodic oxide as a mask, impurities (for example, phosphorous (P)) are driven by using a method such as ion implantation or ion doping to form impurity doped regions 207. (FIG. 2(B))
Thereafter, high-intensity light such as laser light is radiated from above to activate the regions where the impurities have been introduced. (FIG. 2(C))
Finally, an inter-layer insulator 208 is deposited, contact holes are opened over the doped regions, and electrodes 209 are formed for connection to the doped regions, thus completing the fabrication of the thin-film transistor. (FIG. 2(D))
However, it has been found that in the above-described process, the boundaries (indicated by X in FIG. 2(C)) between the doped regions and the active region (the semiconductor region directly below the gate and flanked by the doped regions) are unstable, and that the reliability decreases due to increased leakage current, etc. after use for long periods of time. That is, as can be seen from the process, the crystallinity of the active region remains substantially unchanged throughout the process; on the other hand, the doped regions adjacent to the active region initially have the same crystallinity as that of the active region but their crystallinity is damaged during the process of impurity introduction. The doped regions are repaired in the subsequent laser radiation step, but it is difficult to restore the original crystallinity. Furthermore, it has been found that in particular, the portions of the doped regions that contact the active region cannot be activated sufficiently as such portions tend to remain unexposed to laser radiation. This results in discontinuity in the crystallinity between the doped regions and the active region, tending to cause trapping states, etc. In particular, when impurities are introduced using a method that involves driving high-velocity ions, the impurity ions are caused to scatter and penetrate into regions below the gate electrode, so that the crystallinity of these regions is damaged. It has not been possible to activate such regions lying below the gate electrode by laser or other light since they are in the shadow of the gate electrode.
One way to solve this problem is to radiate laser or other light from the reverse side to activate these regions. In this method, the boundaries between the active region and the doped regions can be activated sufficiently since the light is not blocked by the gate connection. This method, however, requires that the substrate material be transparent to light, and as a matter of course, cannot be employed when a silicon wafer or the like is used as the substrate. Furthermore, most glass materials do not easily transmit ultraviolet light of wavelength below 300 nm; therefore, KrF excimer lasers (wavelength 248 nm), for example, that achieve excellent mass productivity cannot be used.